fix(langchain): handle ToolNode object in create_react_agent instrumentation

[SanjanaB123]

Summary

• Fixes 🐛 Bug Report: TypeError: 'ToolNode' object is not iterable in patch.py when using langgraph-supervisor with create_react_agent #3841
• ToolNode (from langgraph-prebuilt) is not iterable, but patch.py assumed tools was always a list
• Added hasattr(tools, 'tools_by_name') check to extract tools via tools_by_name.values() when a ToolNode is passed
• Added regression test test_create_react_agent_with_toolnode

Reproduction

from langgraph.prebuilt import ToolNode
from langchain_core.tools import tool

@tool
def get_weather(city: str) -> str:
    """Get weather for a city"""
    return "Sunny"

tool_node = ToolNode([get_weather])
for t in tool_node:  # TypeError: 'ToolNode' object is not iterable
    print(t)

• I have added tests that cover my changes.
• If adding a new instrumentation or changing an existing one, I've added screenshots from some observability platform showing the change.
• PR name follows conventional commits format: feat(instrumentation): ... or fix(instrumentation): ....
• (If applicable) I have updated the documentation accordingly.

Summary by CodeRabbit

• Bug Fixes

  • Improved tool-definition handling to support alternative tool container types, preventing a TypeError and ensuring GenAI tool definitions are properly captured in telemetry.

• Tests

  • Added a regression test to validate agent creation with the alternative tool container and to verify semantic telemetry attributes are recorded.

[coderabbitai]

📝 Walkthrough

Walkthrough

This change updates the agent wrapper to handle LangGraph ToolNode inputs by iterating tools.tools_by_name.values() when present, and adds a regression test confirming create_react_agent accepts ToolNode instances and records expected GenAI semantic attributes.

Changes

Cohort / File(s)	Summary
Tool iteration logic fix packages/opentelemetry-instrumentation-langchain/opentelemetry/instrumentation/langchain/patch.py	Adjusts tool-extraction to detect tools_by_name on tools and iterate tools.tools_by_name.values() when available; falls back to previous iterable behavior otherwise.
ToolNode integration test packages/opentelemetry-instrumentation-langchain/tests/test_langgraph.py	Adds test_create_react_agent_with_toolnode to validate ToolNode can be passed to create_react_agent and that GenAI attributes (operation, agent name, and GEN_AI_TOOL_DEFINITIONS containing get_weather and get_time) are set. Minor docstring/formatting tweaks in an existing mock.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Poem

🐰 I nibbled at code with curious paws,
Found a ToolNode with iterable flaws,
I checked for tools_by_name, then hopped through the list,
Now agents run smooth — a rabbit-approved fix! ✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 71.43% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically describes the main fix: handling ToolNode objects in create_react_agent instrumentation, which directly addresses the changeset.
Linked Issues check	✅ Passed	The PR successfully implements the exact fix specified in issue #3841: checking for tools_by_name attribute and iterating tools.tools_by_name.values() when present, with comprehensive test coverage.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to fixing the ToolNode iteration issue: the patch logic modification and the regression test. No unrelated modifications are present.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

⚔️ Resolve merge conflicts

• Resolve merge conflict in branch fix/langchain-toolnode-iterable

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (2)

packages/opentelemetry-instrumentation-langchain/tests/test_langgraph.py (2)

827-838: Consider extracting MockChatModel to reduce duplication.

MockChatModel is defined identically in three tests (test_create_react_agent_span, test_create_agent_with_system_prompt, and this test). Extracting it to a module-level fixture or helper class would reduce maintenance burden.

♻️ Proposed refactor

Add a module-level mock class or fixture at the top of the file:

class MockChatModel(BaseChatModel):
    """Mock chat model for testing agent creation."""

    `@property`
    def _llm_type(self) -> str:
        return "mock"

    def _generate(self, messages, stop=None, run_manager=None, **kwargs):
        return ChatResult(
            generations=[ChatGeneration(message=AIMessage(content="Mock"))]
        )

    def bind_tools(self, tools, **kwargs):
        return self

Then use MockChatModel() directly in each test without redefining it.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/opentelemetry-instrumentation-langchain/tests/test_langgraph.py`
around lines 827 - 838, Extract the duplicate MockChatModel class into a single
module-level helper or fixture so tests reuse it instead of redefining; create
the class named MockChatModel once at top of the file (or as a pytest fixture
returning MockChatModel()), then update each test that currently defines
MockChatModel (test_create_react_agent_span,
test_create_agent_with_system_prompt, and test_langgraph.py's current test) to
instantiate or reference that shared MockChatModel rather than declaring the
class locally.

---

811-871: Well-structured regression test for issue #3841.

The test properly validates that ToolNode objects are handled without raising TypeError, and verifies that tool definitions are correctly extracted via tools_by_name.values().

Consider adding a comment or separate test case for edge cases like an empty ToolNode (i.e., ToolNode([])) to ensure the instrumentation handles that gracefully, though this is optional.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/opentelemetry-instrumentation-langchain/tests/test_langgraph.py`
around lines 811 - 871, Add a new regression test similar to
test_create_react_agent_with_toolnode that constructs an empty ToolNode
(ToolNode([])), calls create_react_agent with model=MockChatModel() and
tools=that empty ToolNode, then assert no exception is raised and that the
created span (find by "create_agent" in spans) contains
GenAIAttributes.GEN_AI_TOOL_DEFINITIONS (which should parse to an empty
list/object); reuse the same MockChatModel, span_exporter usage and JSON-parsing
pattern from test_create_react_agent_with_toolnode to validate the
instrumentation handles empty ToolNode gracefully.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@packages/opentelemetry-instrumentation-langchain/tests/test_langgraph.py`:
- Around line 827-838: Extract the duplicate MockChatModel class into a single
module-level helper or fixture so tests reuse it instead of redefining; create
the class named MockChatModel once at top of the file (or as a pytest fixture
returning MockChatModel()), then update each test that currently defines
MockChatModel (test_create_react_agent_span,
test_create_agent_with_system_prompt, and test_langgraph.py's current test) to
instantiate or reference that shared MockChatModel rather than declaring the
class locally.
- Around line 811-871: Add a new regression test similar to
test_create_react_agent_with_toolnode that constructs an empty ToolNode
(ToolNode([])), calls create_react_agent with model=MockChatModel() and
tools=that empty ToolNode, then assert no exception is raised and that the
created span (find by "create_agent" in spans) contains
GenAIAttributes.GEN_AI_TOOL_DEFINITIONS (which should parse to an empty
list/object); reuse the same MockChatModel, span_exporter usage and JSON-parsing
pattern from test_create_react_agent_with_toolnode to validate the
instrumentation handles empty ToolNode gracefully.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 3bd08649-87a6-4003-a96c-847d60eaa0e7

📥 Commits

Reviewing files that changed from the base of the PR and between 33ca6da and 2500817.

⛔ Files ignored due to path filters (1)

• packages/opentelemetry-instrumentation-langchain/uv.lock is excluded by !**/*.lock

📒 Files selected for processing (1)

• packages/opentelemetry-instrumentation-langchain/tests/test_langgraph.py